shabda 0.1.0

A semantic audio samples curator for livecoding software

Shabda

Shabda is a tool to fetch random samples from https://freesound.org/ based on given words, for use in impro sessions on instruments such as Tidal Cycles and Estuary.

Shabda is the Sanskrit word for "speech sound". In Sanskrit grammar, the term refers to an utterance in the sense of linguistic performance.

Install

• Install Python 3: https://www.python.org/
• Install pip: https://pypi.org/project/pip/
• Install pipenv: pip install pipenv
• Install dependencies: pipenv install
• Install ffmpeg: https://ffmpeg.org/ (e.g. Debian/Ubuntu: apt install ffmpeg)

Use (command line)

In order to download a sample pack, execute in the terminal pipenv run python shabda_cli.py <definition> --licenses <license_name>.

Any word can be a pack definition. If you want more than one sample, separate words by a comma: blue,red

You can define how many variations of a sample to assemble by adding a colon and a number. e.g. blue,red:3,yellow:2 will produce one 'blue' sample, three 'red' samples and two 'yellow' sample.

The optional --licenses parameter allows to fetch only samples that have the specified license. Multiple licenses can be allowed by repeating the --licenses argument. Possible licenses are cc0 (Creative Commons Zero), by (Creative Commons Attribution), and by-nc (Creative Commons Attribution Non-Commercial).

Full example:

pipenv run python shabda_cli.py spaghetti:2,monster:4 --licenses cc0 --licenses by

Use (web application)

Launch the web application:

In debug mode:

FLASK_APP=shabda FLASK_DEBUG=1 pipenv run flask run

In production:

pipenv run gunicorn --workers=4 "shabda:create_app()" -b localhost:8000

Test

pipenv run pytest

Notes

With Estuary, Shabda makes use of this feature: https://github.com/dktr0/estuary/wiki#adding-sound-files-to-estuarywebdirt-on-the-fly

To do

• Move all this to github issues or project :)
• Explain how to launch on codespace / how to make port public
• Change allowed duration in definition ?
• List API and view
• Solidify: Retry, Queue workers, queue status?
  • https://huey.readthedocs.io/en/latest/index.html
  • https://github.com/litements/litequeue
  • (but with those there is no way to know a queue's status?)
• Explain risks about throttling: https://freesound.org/docs/api/overview.html
• Allow to rename sample ?
  • Allow emojis in sample name ?
  • Allow spaces in sample name ? special unicode spaces ? Reverse ?
• Record your voice
• Extract percussive ?
  • https://librosa.org/doc/latest/generated/librosa.effects.hpss.html
• Use advanced search in a funky way: https://freesound.org/docs/api/analysis_docs.html#analysis-docs
• Fix refresh token method. Refresh regularly ?
• Enforce all alpha minuscules and numbers in pack definition
• Generate speech ;
  • via external cli ? https://linuxhint.com/command-line-text-speech-apps-linux/
  • via python lib ? https://pypi.org/project/pyttsx3/
  • via TTS API? https://rapidapi.com/collection/best-text-to-speech-apis
• Delete old samples (e.g. 7 days after last usage, note usage date)
• Autorun pylint
• Autorun pytest
• Better cli interface (pack definition, licenses)
• pip module
  • Separate web/cli/core
  • Finish migration from pipenv to Poetry
  • Store freesound token in a predictable directory
  • Store samples in a predictable directory (configurable)